Updating multivariate calibrations with the delaunay triangulation method.

Area of Science:

• Analytical Chemistry
• Chemometrics
• Data Science

Background:

• Multivariate calibration models require updates when new samples exhibit distinct spectral properties.
• Existing methods for updating calibration sets can be complex and computationally intensive.

Purpose of the Study:

• To present and evaluate an updating procedure for multivariate calibration models using Delaunay triangulation (DT).
• To compare the performance of the DT updating method against partial least squares (PLS) for predictive accuracy.

Main Methods:

• Detection and classification of outliers in new samples into marginal and true outliers.
• Updating the calibration set with marginal outliers or forming new local models with true outliers.
• Topological multivariate calibration based on Delaunay triangulation (DT).

Main Results:

• The proposed DT updating procedure effectively handles new samples with different spectral characteristics.
• Prediction results using the updated DT method were comparable to or better than those obtained with PLS.
• The method successfully identified and utilized outliers for model refinement or creation of local models.

Conclusions:

• The Delaunay triangulation (DT) method provides a valuable and efficient approach for updating multivariate calibration models.
• DT-based updating is a robust alternative to traditional methods like PLS, especially when dealing with evolving data.
• The strategy of distinguishing and utilizing marginal and true outliers enhances model adaptability and predictive power.